The breakout AI role, decoded

Forward deployed engineer: the breakout AI role, explained

What is a forward deployed engineer?

A forward deployed engineer is a software engineer who works inside the customer's environment rather than back at headquarters building one feature for everyone. Where a traditional engineer creates a single capability used by many customers, an FDE enables many capabilities for a single customer — analysing the real problem, designing the integration, and shipping production code on the customer's infrastructure. Palantir's own framing is blunt: the responsibilities look like those of a startup CTO — small teams, end-to-end ownership of high-stakes projects.

The "forward deployed" name is borrowed from the military sense of being stationed in the field. FDEs typically split time between onsite customer work and their home engineering org, and — crucially — they feed what they learn in the field back into the core product. That feedback loop is the difference between an FDE and a one-off consultant: the FDE makes the platform better, not just the deployment.

Why the role is exploding now

The FDE pattern is roughly fifteen years old, but demand spiked because of a specific gap in AI adoption: the model is the easy part, and the deployment is the hard part. Enterprises have sensitive data, bespoke systems, and ambiguous requirements; off-the-shelf AI products stall there. An engineer embedded onsite, building against the real environment and feeding insights back to the product team, is how that gap gets closed.

So the frontier labs built FDE functions. OpenAI launched a forward deployed engineering team in 2025 that grew from two engineers to more than ten. Anthropic's Applied AI team hires FDEs to embed with strategic customers and deliver technical artifacts — production applications on Claude, MCP servers, sub-agents — directly in enterprise environments. In May 2026 both labs went further, announcing large deployment ventures within days of each other; Anthropic's is a $1.5B joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs. Venture investors have taken to calling FDE "the hottest job in tech." The signal for a senior engineer: this is where the demand, and the comp, are heading.

FDE vs AI engineer vs solutions / sales engineer

The titles overlap and aren't standardised, but the useful axis is where you work and what you ship. An FDE ships production code on customer infrastructure; a solutions or sales engineer mostly builds demos and proofs-of-concept to support a sale. An AI engineer builds and operates the product back in the core org. The FDE sits deliberately in between — production engineer with customer proximity.

For a fuller breakdown of how these titles collapse in real postings, see the AI roles, decoded. If you'd rather own the design than the deployment, the AI architect role is the adjacent senior path. The skill set is ~70% shared across all of them; pick a lean, not a different education.

What an FDE actually does day to day

The work is end-to-end ownership under real-world ambiguity. A typical engagement runs from discovery to production: understand the customer's actual problem (rarely the one in the brief), design an integration, build it on their stack, and harden the prototype into something secure, compliant, and operable. Onsite time is common — often a meaningful share of the role — and the through-line is shipping working software, not slideware.

• Discovery in the fog. Client environments rarely come with clean requirements; the first job is decomposing an ambiguous problem into something buildable.
• Build on customer infrastructure. Integrations, data pipelines, auth flows, retries, rate limits, schema evolution — production concerns, not demo concerns.
• Harden to production. Turn the working prototype into a secure, compliant, scalable system that survives real use.
• Close the loop. Carry field learnings back to the core product so the next deployment is easier.

A commonly cited 90-day ramp captures the shape: days 1–30, learn the product and ship a small customer win; days 31–60, own a deployment end-to-end and contribute reusable integrations; days 61–90, drive cross-customer improvements informed by what you saw in the field.

Skills you need

FDE is a full-stack engineering role with a customer-facing edge, and — increasingly — an AI-native layer on top:

• Strong general engineering — Python is the lingua franca; TypeScript / JavaScript, Go, or Java for full-stack delivery.
• Data and systems fluency — SQL beyond the basics, data pipelines and processing, plus cloud (AWS / GCP / Azure), Docker, and Kubernetes.
• Integration craft — resilient APIs, auth flows, secure secret storage, retries, rate limits, and schema evolution.
• The AI-native layer — LLM and agentic fundamentals: tool use, MCP, sub-agents, retrieval, and evals so you can measure correctness when outputs vary. See designing agentic AI systems.
• Ambiguity tolerance and communication. The differentiator isn't a framework — it's thriving when the requirements are undefined and the stakeholder is in the room.

You do not need machine learning or model training for this work — FDE work is building systems on top of existing models, exactly like AI engineering. Here's why ML isn't a prerequisite.

Salary (directional)

FDE compensation is high and unusually wide, because it spans new-grad through principal and clusters heavily at well-funded AI companies. Treat the figures below as directional — they're drawn from 2026 public salary aggregators and compensation reports whose methodologies and title definitions vary.

The headline: even the broad-market median sits comfortably above general software-engineering pay, and the frontier-lab band is among the highest in applied AI. The spread is real, though — what you earn depends heavily on company stage, level, and location, so read any single number as a marker, not a promise.

How to become an FDE (for senior engineers)

Because the audience is already senior, the transition is measured in weeks of focused building, not years of study. The role rewards engineers who can ship, decompose ambiguity, and talk to a customer — so the path is about evidence, not credentials:

• 1. Sharpen the full-stack core. Be fluent in Python plus one of TypeScript / Go / Java, and comfortable across data, cloud, and containers.
• 2. Build a real agent end to end. A tool-calling agent with retrieval and evals — production-shaped, not a notebook demo.
• 3. Make it deployable on someone else's stack. Integrations, auth, secrets, retries, observability, and cost control — the things that break in the field.
• 4. Practise decomposition. FDE interviews famously hinge on an open-ended case: an ambiguous, real-world problem with no single right answer. Clarify, break it into chunks, propose a simple MVP, then iterate — out loud.
• 5. Ship a capstone with a write-up. A runnable system plus the decisions behind it — the artifact that proves you can deliver in the fog.

Where you start shapes the fastest route. From a DevOps or platform background, your operations moat — observability, cost, reliability — is exactly what makes a deployment survive contact with a customer.

Frequently asked questions

What is a forward deployed engineer?

A forward deployed engineer is a full-stack software engineer who embeds directly with a customer and builds production systems inside their environment, rather than building one feature for everyone back at headquarters. The role started at Palantir (internally "Delta") and is now central to how OpenAI and Anthropic ship AI into enterprises.

What does a forward deployed engineer do?

They own a customer engagement end to end: discover the real problem, design an integration, build it on the customer's infrastructure, and harden the prototype into a secure, compliant, production system — then feed what they learned back into the core product. Onsite time is common, and the deliverable is working software, not a demo.

How much does a forward deployed engineer make?

Directionally, 2026 public aggregators put the broad US median around $183k–$190k, with entry roles roughly $180k–$250k at well-funded companies and frontier labs such as OpenAI and Anthropic in the $350k–$550k band for mid-to-senior engineers; staff and principal can exceed $600k. These figures vary by methodology, level, company stage, and location, so treat them as markers.

Is forward deployed engineer a good role?

For an engineer who likes shipping and being close to the problem, yes — it's well paid, in high demand, and sits at the centre of how enterprises actually adopt AI. The trade-offs are real: ambiguity, customer-facing pressure, and often travel. If you prefer steady-state product work over field delivery, a core AI engineering role may fit better.

How is an FDE different from a software engineer?

A traditional software engineer builds one capability used by many customers, back in the core org. An FDE enables many capabilities for a single customer, working inside that customer's environment and shipping production code on their infrastructure — then carrying field insight back to product. Same engineering rigour, different centre of gravity.

Do you need machine learning to be an FDE?

No. FDE work — including forward deployed AI engineering — is about building systems on top of existing models: retrieval, tool use, MCP, sub-agents, evals, and integration. You need strong general engineering and judgement, not model training. ML research is a separate discipline.

How do you become a forward deployed engineer?

If you're already a senior engineer, sharpen your full-stack core, build a real tool-calling agent with evals, make it deployable on someone else's stack, practise decomposing ambiguous problems out loud (the interview hinges on it), and ship a capstone with a written rationale. It's weeks of focused building, not years of study.